open source pkg v1

pkg/OpenFace/matlab_runners/Demos/run_demo_images.m

@@ -0,0 +1,34 @@
+clear
+if(isunix)
+    executable = '"../../build/bin/FaceLandmarkImg"';
+else
+    executable = '"../../x64/Release/FaceLandmarkImg.exe"';
+end
+    
+in_dir  = '../../samples/';
+out_dir = './demo_img/';
+
+model = 'model/main_ceclm_general.txt'; % Trained on in the wild, menpo and multi-pie data (a CE-CLM model)
+
+% Uncomment the below models if you want to try them
+%model = 'model/main_clnf_general.txt'; % Trained on in the wild and multi-pie data (a CLNF model)
+
+%model = 'model/main_clnf_wild.txt'; % Trained on in-the-wild data only
+
+%model = 'model/main_clm_general.txt'; % Trained on in the wild and multi-pie data (less accurate SVR/CLM model)
+%model = 'model/main_clm_wild.txt'; % Trained on in-the-wild
+
+% Load images (-fdir), output images and all the features (-out_dir), use a
+% user specified model (-mloc), and visualize everything (-verbose)
+command = sprintf('%s -fdir "%s" -out_dir "%s" -verbose -mloc "%s" ', executable, in_dir, out_dir, model);
+
+% Demonstrates the multi-hypothesis slow landmark detection (more accurate
+% when dealing with non-frontal faces and less accurate face detections)
+% Comment to skip this functionality
+command = cat(2, command, ' -wild -multi_view 1');
+
+if(isunix)
+    unix(command);
+else
+    dos(command);
+end